The smart city concept has been popularized in the urbanization of major metropolitan areas through the implementation of intelligent systems and technology to serve the increasing human population. This work developed an automatic light adjustment system at Thammasat University, Rangsit Campus, Thailand, with a primary objective of optimizing energy efficiency, while providing sufficient illumination for the campus. The development consists of two sections: the device control and the prediction model. The device control functionalities were developed with the user interface to enable control of the smart street light devices and the application programming interface (API) to send the light-adjusting command. The prediction model was created using an AI-assisted data analytic platform to obtain the predicted illuminance values so as to, subsequently, suggest light-dimming values according to the current environment. Four machine-learning models were performed on a nine-month environmental dataset to acquire predictions. The result demonstrated that the three-day window size setting with the XGBoost model yielded the best performance, attaining the correlation coefficient value of 0.922, showing a linear relationship between actual and predicted illuminance values using the test dataset. The prediction retrieval API was established and connected to the device control API, which later created an automated system that operated at a 20-min interval. This allowed real-time feedback to automatically adjust the smart street lighting devices through the purpose-designed data analytics features.
This work tackles audio signal analysis problems related to snoring sounds of simple and obstructive sleep apnea/hypopnea syndrome (OSAHS) snorers by optimizing acoustic feature extraction and training various machine learning models. Relevant acoustic features, including Mel-frequency cepstral coefficients, zero crossing rate, root mean square, and spectral features, were extracted from snoring sound recordings. Two main tasks are performed: firstly, the classification of snoring sounds of simple snorers versus OSAHS patients, and secondly, the discrimination between apnea/hypopnea versus normal snoring events of OSAHS patients. The results show an outstanding performance in differentiating between the snoring sounds of simple snorers and OSAHS patients with an accuracy of 0.9982. The challenge is upon the binary classification between apnea/hypopnea versus normal snoring events in patients with OSAHS, which was evaluated using the subject-independent leave-one-subject-out cross-validation (LOSOCV) technique, yielding the best accuracy of 0.7838 and an F1-score of 0.7734, with the random forest model. Energy-related acoustic features were found to be crucial in distinguishing snoring patterns. In addition, the apnea/hypopnea index (AHI) was estimated from the snoring pattern classification using the LOSOCV technique. The extra trees classifier demonstrated a 0.8689 Pearson correlation coefficient (PCC) between predicted and PSG-measured AHI values. The highest performance in predicting three groups of AHI severity (mild, moderate, and severe), based on each subject’s AHI value, was achieved by the decision trees model with an accuracy of approximately 0.8621. Voting classifier, constructed from five classifiers, provides reliability of prediction by reducing bias through leveraging multiple models to mitigate errors inherent in binary classification, achieving an accuracy of 0.8276. The study proposes an estimation of AHI severity levels by analyzing individual events of snoring sounds, paving the way for applying acoustic feature analysis for unbiased, simple, and accurate pre-screening of sleep-related breathing disorders.
Somrudee Deepaisarn, Paphana Yiwsiw, Chanon Tantiwattanapaibul, Suphachok Buaruk, and Virach Sornlertlamvanich. Journal of information and communication convergence engineering 2023;21:216-24. https://doi.org/10.56977/jicce.2023.21.3.216